Drug‐Induced Liver Injury in Patients With Chronic Liver Disease

ABSTRACT

Objective

Drug‐induced liver injury (DILI) is a global problem and can develop from exposure to prescription or over‐the‐counter medications as well as herbal and dietary supplements. The diagnosis of DILI is clinically challenging, and liver injury can be severe leading to liver failure, death, or liver transplantation. Patients with underlying chronic liver diseases (CLD) may be at increased risk for DILI, which is associated with factors related to drug or liver disease.

Methods

This review summarises current knowledge on the risk and outcomes of DILI in patients with CLD.

Results

Patients with CLD may be at an increased risk for DILI. Additionally patients with underlying CLD are at risk for more severe liver injury and worse outcomes after DILI.

Discussion

The risk for and poor outcomes from DILI are accentuated in patients with CLD and potentially leading to the worst‐case scenario of acute‐on‐chronic liver failure. We highlight the key observations on DILI with a broad range of underlying liver diseases and the high‐DILI risk agents implicated in those populations.

Abbreviations

AIH

autoimmune hepatitis

APAP

acetaminophen

CLD

chronic liver disease

DILI

drug‐induced liver injury

DILIN

drug‐induced liver injury network

ER

endoplasmic reticulum

HBV

hepatitis B virus

HCV

hepatitis C virus

HDS

herbal and dietary supplements

IRR

incidence rate ratios

LT

liver transplantation

MASH

metabolic‐dysfunction associated steatohepatitis.

MASLD

metabolic‐dysfunction associated steatotic liver disease

MELD

model for end‐stage liver disease

PBC

primary biliary cholangitis

PPAR

peroxisome proliferator‐activated receptor

RUCAM

The Roussel Uclaf Causality Assessment Method

TB

tuberculosis

TZD

thiazolidinedione

ULN

upper limit of normal

Summary.

• Patients with underlying chronic liver disease (CLD) appear to be at risk for DILI.

• The strongest evidence for this association is in patients with viral liver disease and MASLD.

• CLD is an important consideration when prescribing agents with a relatively strong association with DILI when reasonable and less risky alternative agents are on hand.

• Recognising the presence of underlying chronic liver disease in patients with DILI is important for prognostic implications.

• Patients with CLD and DILI are at risk for more severe liver injury.

• Patients with CLD and DILI experience worse outcomes including mortality and need for liver transplantation.

1. Introduction

The late Hyman Zimmerman, in his classic monograph, wrote, ‘a stubborn misconception regarding susceptibility to hepatic injury has been the view that patients with preexisting liver disease are more likely than others to experience a hepatic injury on exposure to drugs that cause liver damage’ [1]. He further writes that there is virtually no evidence for this view other than isolated examples. Since that writing a quarter century ago, our understanding of pre‐existing liver disease as a risk factor for idiosyncratic drug‐induced liver injury (DILI) and its outcomes is still evolving. In this review article, we summarise recent developments in this space. While it is clear that DILI occurring in patients with underlying liver disease is associated with worse outcomes [2, 3], the relationship between underlying liver disease and susceptibility to DILI is likely more nuanced and may be agent and type of underlying liver disease‐specific.

2. The Impact of Underlying Liver Disease on Risk of Drug‐Induced Liver Injury

Chronic liver disease (CLD) may contribute to the risk of DILI through multiple mechanisms, but this association is difficult to study due to the infrequency of DILI with any individual drug exposure. This may be further confounded by heterogeneity in CLD etiologies and the type of drug exposure impacted by medical and geographic considerations. These will be discussed in relation to categories of underlying CLD, with the most compelling data in patients with underlying viral liver disease and metabolic dysfunction‐associated steatotic liver disease (MASLD).

Patients with underlying liver disease may be at increased risk for DILI due to the liver's essential role in drug metabolism. In addition to a decrease in the rate of drug metabolism from a decrease in drug‐metabolising enzymes, the decreased synthetic function with low albumin levels may also result in increased unbound drug levels [4, 5, 6, 7]. Furthermore, patients with portal hypertension and collateral circulation may have altered drug disposition and pharmacokinetics from altered hepatic blood flow, leading to changes in first‐pass metabolism [4, 5, 6]. Additionally, individuals with CLD often take multiple medications, increasing the risk of drug–drug interactions and, consequently, a higher risk of DILI.

2.1. DILI In Individuals With Underlying Viral Liver Disease

Several possible interactions between DILI and some viral liver diseases bear consideration. These include (i) the potential for increased risk of DILI with underlying viral liver disease, (ii) an increased frequency of DILI due to specific drugs and use of HDS in populations with higher rates of endemic viral liver disease, (iii) the potential for viral reactivation to mimic DILI and (iv) the potential for viral liver disease to induce secondary liver conditions that may mimic DILI, such as autoimmune hepatitis (AIH).

Chronic viral hepatitis, such as hepatitis B (HBV) and hepatitis C (HCV), are known to induce endoplasmic reticulum (ER) stress [8, 9], oxidative stress [10], and mitochondrial dysfunction [11, 12]. These cellular stresses are important factors in the pathogenesis of viral hepatitis and disease progression but, theoretically, they may also provide a milieu to lower the threshold for DILI. The risk factors for intrinsic or dose‐dependent DILI include the interrelated processes of ER and oxidative stress, mitochondrial dysfunction, and alterations in bile acid haemostasis [13]. These same mechanisms are also implicated in idiosyncratic, or non‐dose dependent, DILI [13, 14], as supported by cell‐based assays of idiosyncratic DILI liability using hepatocyte/hepatocyte‐like cell lines [15, 16]. These data suggest that stress responses probably occur to some degree in a large subset of drug‐exposed individuals, though clearly, additional factors are needed for rare but clinically apparent idiosyncratic DILI to develop, and clear evidence of a potential association is lacking [13].

Direct clinical evidence supporting the association of viral hepatitis and DILI is derived indirectly from a few animal studies. In the absence of an animal model of idiosyncratic DILI, one may consider animal studies on intrinsic hepatotoxicity as seen in acetaminophen (APAP) liver injury. In a mouse model, both viral and bacterial‐induced inflammation increased the sensitivity of animals to APAP hepatotoxicity, which occurred at lower exposures [17]. However, conflicting data demonstrate decreased APAP hepatotoxicity with adenoviral and retroviral infection in mice [18, 19]. The primary evidence supporting an increased risk of DILI with underlying viral liver disease derives from data in patients with HBV and HCV infection and undergoing treatment for tuberculosis (TB).

In individuals with chronic viral hepatitis, the immune system plays a crucial role in controlling viral replication and associated inflammation through a coordinated response involving innate and adaptive immune systems. This delicate balance between viral replication and immune surveillance helps to prevent the progression of viral hepatitis. However, when medications induce immune suppression, there is a risk of viral reactivation, leading to a hepatitis flare [20]. This form of liver injury also termed indirect DILI, results from the medication's impact on the immune system rather than from direct hepatotoxicity [20]. The injury may manifest as a worsening of preexisting viral hepatitis or the onset of new liver disease, with clinical symptoms that mimic those of the underlying viral condition [20].

Aminotransferase flares, sometimes leading to fulminant hepatic failure, were initially reported in patients with chronic HBV who received chemotherapy for a variety of malignancies who did not receive pretreatment prophylaxis with HBV therapy [21]. It is currently the standard of care to check for HBV serology before initiating chemotherapy or a biologic known to cause immune suppression [22]. For patients with chronic HBV or evidence of prior exposure (e.g. core antibody positive), prophylactic antiviral therapy is recommended before chemotherapy and continued for at least six months after its completion [22]. The risk of viral disease flare is highest after completion of chemotherapy due to immune reconstitution. The risk of flare from chronic HCV presenting as a DILI is not well characterised but can occur in 23% of treated patients [23, 24]. Unlike anticancer chemotherapy, which can indirectly trigger acute hepatitis by reactivating HBV, antiretroviral treatments may lead to immune reconstitution, exacerbating HCV infection.

Due to the rarity of DILI from any specific drug exposure, it is difficult to establish DILI risk for specific agents or classes of agents with underlying CLD when large numbers are needed to discern differential risk. By example, in the 2015 report from the US DILIN, there was overrepresentation of azithromycin DILI in patients with CLD (6.7% vs. 1.5% without CLD) [2], but to our knowledge this observation has not been reproduced. Further, a recent paper from the US DILIN on azithromycin DILI did not comment about underlying CLD as a risk factor for azithromycin DILI [25]. As a result, it would be imprudent to overemphasise this observation or specifically deny patients with CLD treatment with this commonly prescribed agent.

2.1.1. DILI In Individuals With Underlying Hepatitis B Infection

The prevalence of HBV infection varies globally, with similarly variable rates of other infections such as tuberculosis (TB). With the recognised potential for DILI from TB therapies, isoniazid, rifampicin, and pyrazinamide, patients with HBV and TB coinfection treated with these agents represent an at‐risk group and allows an assessment of the impact of HBV infection on the risk of DILI. A recent systematic review identified the highest pooled prevalence coinfection rates in Africa [11.4%, 95% confidence interval (CI): 3.45–19.31] and the Western Pacific region (10.8%, 95% CI 8.68–12.84) and the lowest rates in the Americas (2.2%, 95% CI 0.78–3.53) [26]. Not surprisingly, most of the data on DILI from TB medications in co‐infected patients derives from Western Pacific populations, with little data from African countries.

A 2016 systematic review of 15 prospective and case–control studies (13 studies included Asian patients, 2 included Caucasian patients) examined DILI from combination TB therapy in 11 and isoniazid therapy in 4 studies [27]. Patients with HBV infection had a higher pooled odds ratio (2.18, 95% CI 1.41–3.37) of developing suspected DILI compared to patients without HBV infection, albeit with heterogeneous definitions of DILI (ranging from any ALT elevation to ALT increase > 5 times upper limit of normal (ULN)). In subgroup analyses of prospective studies with a strict definition of DILI ALT increase > 5 × ULN the pooled odds ratio of DILI with HBV infection was 3.41 (95% CI: 1.77–6.59) with combination TB therapy (4 studies), but was not significant (3.8, 95% CI: 0.31–467) for isoniazid therapy (2 studies).

A more recent meta‐analysis assessed 10 prospective and retrospective studies examining combination TB therapy (excluding patients with underlying cirrhosis or studies with unclear definitions of DILI), thst is, uncertain consistency, and impact of causality adjudication among the included studies. It described similar elevations in the risk ratio of DILI in patients with HBV infection (1.98, 95% CI: 1.38–2.83) relative to those without HBV infection [28]. The risk ratio was similar in subgroup analyses of prospective studies (2.29, 95% CI: 1.79–2.90) or studies published after 2000 (2.75, 95% CI: 2.1–3.59). Another meta‐analysis of 16 studies described a higher risk ratio of DILI in TB and HBV co‐infected patients (2.66, 95% CI: 2.13–3.32), but a higher risk ratio in those who are hepatitis B E antigen positive (3.42, 95% CI: 1.95–5.98), and a longer duration of recovery from DILI compared to non‐ coinfected patients [29].

The mechanism of DILI in the setting of TB therapy in HBV coinfection is not well‐understood, although it may involve the aforementioned mechanisms, altered pharmacokinetics and dynamics due to CLD, and possibly HBV flares. Interestingly, the risk of DILI from TB therapies in coinfected patients can be attenuated with antiviral therapy for hepatitis B. In a propensity score matched population‐level study in Hong Kong, patients on HBV therapy for more than 1 year before receiving TB therapy had a lower risk of hospitalisation due to DILI (adjusted hazard ratio, 0.44; 95% confidence interval 0.26–0.72) [30]. Similar trends were reported in HBV‐HIV coinfected patients receiving treatment for HBV and HIV ahead of TB therapy [31]. However, it is conceivable that at least some of the patients considered to have DILI, might have had flares of HBV in these studies.

Another impact of the global variability in HBV is the associated variability in cultural norms relating to HDS and medicinal herbal therapies and the DILI that may be associated with their use. As an example, the use of traditional Chinese medicines is superimposed on the higher prevalence of HBV in the Chinese population. There are no compelling data to estimate the risk of DILI from these herbal agents in those with underlying HBV versus those who do not, and the only data available derives from reports of acute‐on‐chronic liver failure (ACLF) from Southeast Asia, which are discussed in the section on outcomes of DILI.

Reactivation of HBV is common in immunosuppressed patients, including those with underlying malignancy and autoimmune conditions, and organ transplant recipients, with a higher risk in patients who are surface antigen positive. Within the context of this review, such flares may mimic DILI, and HBV reactivation has to be considered in any patient with known underlying HBV infection or in non‐viremic patients with detectable hepatitis B core antibody, particularly if not receiving prophylactic antiviral therapy.

AIH frequently confounds the diagnosis of DILI, regardless of precipitating factors [32]. While rare, AIH has been described in patients with clinically evident or occult HBV infection due to loss of immune tolerance and may be associated with delays in diagnosis of AIH [33, 34]. Therefore, AIH should be considered in patients with HBV infection and unexplained liver injury, and the combination of characteristic histology and serologies may aid in diagnosis with favourable treatment outcomes [33]. There are no data estimating the risk of DILI in patients with HBV and hepatitis Delta coinfection.

2.1.2. DILI In Individuals With Underlying Hepatitis C Infection

Similar to HBV infection, HCV infection has increased prevalence in patients with TB (7%–11%) [35], and is also associated with an increased risk of TB therapy‐related DILI [36, 37]. In a systematic review of 16 prospective and case–control studies (14 Asian populations, 4 Caucasian and 1 Black) examining combination (12 studies) and isoniazid (4 studies) therapy, the pooled odds ratio for DILI (defined as ALT > 2–5 × ULN) with chronic HCV infection was 3.21 (95% CI: 2.30–4.49) [37]. The pooled odds ratio for DILI with chronic HCV was significantly increased for Asian (2.96, 95% CI: 1.79–4.90), Caucasian (4.07, 95% CI: 2.70–6.14), and Black (3.25, 95% CI: 2.30–4.60) patients. Similar pooled odds ratios were noted for combination (2.94, 95% CI: 1.95–4.4) and isoniazid (4.18, 95% CI: 2.36–7.40) therapy, and in the analysis of using a strict definition of DILI (ALT greater than 5 × ULN) (2.59, 95% CI: 1.58–4.25). The suspected mechanisms of suspected DILI in HCV‐TB coinfection are similar to those proposed in patients with HBV‐TB coinfection and include the consideration of viral flares [37]. A systematic review on the role of co‐treatment of HCV and TB in coinfected patients suggested that direct‐acting agents for HCV ameliorated suspected DILI in patients with multidrug‐resistant TB on TB therapy [38].

It is important to recognise that acute hepatitis C can mimic hepatocellular DILI, as observed in the DILIN [39]. Additionally, like HBV infection, flares of HCV infection can occur during periods of immunosuppression, and HCV infection is known to be rarely associated with AIH overlap in a subset of patients, resulting in a number of scenarios of acute liver injury, which could mimic or confound a presentation and diagnosis of DILI. Similar considerations have to be made for hepatitis E testing in cases of suspected DILI, particularly in patients presenting with a hepatocellular liver injury pattern [40].

2.2. DILI In Individuals With Suspected Metabolic Dysfunction‐Association Steatotic Liver Disease

MASLD is among the most common liver diseases worldwide and in many individuals it is occult. There is an immense clinical development activity where novel compounds are increasingly being tested to treat metabolic dysfunction‐associated steatohepatitis (MASH) and related cirrhosis. Thus, there is a great need to better understand if MASLD increases the risk for DILI. MASLD is associated with significant changes in drug‐metabolising enzymes and drug transporters [41, 42, 43], although there isn't evidence to link such altered pharmacokinetics per se to increased risk for DILI.

Statins are extensively used worldwide for treating dyslipidemia and cardiovascular disease. Although serious liver toxicity is quite rare, liver enzyme fluctuations are common with statins, and there has been an interest to investigate if underlying liver disease is a risk factor for statin DILI. In a highly cited paper [44], one of the authors of this review article, examined the incidence of DILI from statins in patients with elevated baseline liver enzymes in a pharmaco‐epidemiological study. Their study observed that patients with elevated liver enzymes receiving statins did not have a higher incidence of mild to moderate (4.7% vs. 6.4%, p = 0.2) or severe elevations (0.6% vs. 0.4%, p = 0.6) when compared to individuals with elevated liver enzymes who did not receive statins. This study was followed by a number of studies in other liver disease populations (e.g. hepatitis C), which consistently reiterated the safety of statins in patients with chronic liver diseases such as MASLD and hepatitis C [45, 46, 47]. In fact, statins are being investigated to reduce the risk of decompensation in two multicenter randomised controlled studies in the United States [48, 49]. While we feel statins are very safe in patients with cirrhosis, we advise against their use in patients with decompensated cirrhosis as their significantly altered pharmacokinetics in patients with decompensated cirrhosis may place them at risk for muscle injury and subsequent acute kidney injury [50].

Troglitazone, a first‐generation thiazolidinedione (TZD), was removed from the market because of its hepatotoxicity, and thus, there was hesitation to use second‐generation TZDs in patients with liver disease. Studies published by the authors of this review article demonstrated that lovastatin and rosiglitazone are not associated with the risk of hepatotoxicity in patients with underlying suspected MASLD [51, 52].

While the aforementioned studies indicate statins and TZDs can be safely used in patients with MASLD, other sets of data suggest that MASLD may be a risk factor for DILI. In a prospective study reported from a tertiary care center in Italy in 2007 [53], Tarantino et al. observed that patients with underlying MASLD increased the risk of DILI by nearly 4‐fold, in comparison to a matched cohort of patients with hepatitis C. Noteworthy is that this study defined DILI as ALT > 10 times in the absence of a competing etiologies. Implicated agents in this paper include fosinopril, losartan, piperacillin‐tazobactam plus non‐steroidal anti‐inflammatory drugs, ticlopidine, telithromycin, and omeprazole. In a more recent pharmaco‐epidemiological study [54], we examined the risk of suspected DILI from one of top 10 select medications most frequently implicated to cause DILI in the United States in individuals with underlying suspected MASLD. Suspected DILI was defined as serum ALT > 200 U/L and/or serum alkaline phosphatase > 250 U/L and/or total bilirubin > 2.5 mg/dL on at least two consecutive occasions within 3 months after receiving one of ten candidate prescription medications, in the absence of positive hepatitis B surface antigen, positive HCV antibody, significant alcohol consumption, or hypotension [2]. The incidence of suspected DILI in individuals with underlying suspected MASLD was 0.8%, significantly higher than in individuals with no evidence for MASLD at baseline (0.2%, p < 0.001). This study suffers from the drawback that there was no control group of suspected MASLD patients who did not receive one of these 10 prescription agents. In a subsequent South Korean population‐based database study, Hwang and colleagues reported evidence of an increase in the risk of DILI associated with MASLD [55]. In the study, the risk of developing DILI among patients with MASLD was estimated by the incidence rate ratios (IRR) using baseline period as the reference (> 90 days prior to MASLD diagnosis). The risk of DILI, defined as ICD‐10 code K71, was high within 90 days before the diagnosis of MASLD (IRR, 3.81, p < 0.001), the highest within 7028 days after the diagnosis of MASLD (IRR 7.74, p < 0.001) and decreased thereafter (IRR 1.1, p < 0,001). Collectively, these studies raise the possibility that background MASLD may increase the risk of any‐cause DILI, but higher quality investigations are needed to confirm or refute these observations.

2.3. DILI In Individuals With Underlying Alcohol‐Associated Liver Disease

Alcohol use is a leading risk factor for liver disease globally, although with limited evidence relating to DILI risk with isoniazid, methotrexate and halothane [56]. Alcohol use is considered a risk factor for DILI in the Roussel Uclaf Causality Assessment Method (RUCAM) causality assessment tool with 1 point assigned [57]. Individuals with excess alcohol use and alcohol use disorder are at risk of underlying liver disease with varying severity that may range from steatosis to alcohol‐associated hepatitis and cirrhosis.

Alcohol can interact with prescription drugs in at least two key ways: affecting the pharmacokinetics of drugs or altering how alcohol itself is metabolised in the presence of drugs [58]. Excess alcohol can affect drug metabolism by modifying gastric emptying or inducing drug‐metabolising enzymes such as cytochrome P450 2E1 [59]. The well‐established link between alcohol and acetaminophen toxicity is an example, but alcohol's role in idiosyncratic DILI remains uncertain [60, 61]. Although heavy alcohol use increases the risk of fibrosis and cirrhosis with methotrexate, its impact on other drugs like anti‐TB medications appears to be related to the severity of DILI [62, 63]. Labeling for certain medicines, such as duloxetine, warns against use in heavy alcohol consumers, though evidence of increased hepatotoxicity remains limited [64]. A comprehensive review of the prospective multicenter Drug‐Induced Liver Injury Network (DILIN) study found that anabolic steroids were the most common cause of DILI in heavy consumers of alcohol (2 or more than 2 drink per day in females and 3 or more drinks/day in males) [60]. Beyond that, the liver injury in DILI was not more severe in patients with heavy alcohol use, and outcomes did not differ. In summary, alcohol alters drug metabolism and may pose a higher risk of liver injury with chronic drug use, especially in cases of significant alcohol consumption or pre‐existing liver disease. However, the extent of this risk varies depending on the drug and the pattern of alcohol use.

Given the limited evidence for association of risk, the revised electronic version of RUCAM for the diagnosis of DILI does not include alcohol use, nor age for that matter, as risk factors for DILI [65]. It should be acknowledged that heavy consumption of alcohol is not associated with worse outcomes in patients with DILI compared to non‐drinkers [60]. A common challenge in any study relating to alcohol use is the difficulty in getting a reliable alcohol use history, and studies of DILI incorporating biomarkers of alcohol metabolites may shed more light on any associations.

2.4. DILI In Individuals With Underlying Autoimmune Liver Disease

There is limited data to suggest that patients with autoimmune liver diseases, like AIH, primary biliary cholangitis (PBC), or primary sclerosing cholangitis (PSC), are at an increased risk for DILI. However, it is important to recognise that DILI and autoimmune liver diseases, particularly AIH, can present with similar clinical and histological features, for example, nitrofurantoin DILI can present with autoantibodies and autoimmune hepatitis histologic changes including fibrosis [66]. In addition, DILI can also induce an AIH‐like liver injury that is steroid‐responsive, making it difficult to distinguish DILI from AIH [67]. The diagnostic challenges may be bidirectional since a significant proportion of patients with AIH‐like DILI achieve long‐term remission off immunosuppression, while cases of recurrent DILI often present with features of AIH despite multiple agents being implicated with each ‘DILI’ episode [68].

Patients with PBC may be at risk for DILI, as discussed below in the section on DILI in clinical trials. As in AIH, DILI can be difficult to distinguish from underlying disease activity [69]. It can present with chronic cholestasis and vanishing bile duct syndrome, although the latter condition usually presents more acutely with cholestasis and jaundice and does not improve, but rather progresses to liver failure in severe cases [70]. Less is known about DILI risk in patients with PSC, but DILI presenting with bile duct injury mimicking PSC has been described in up to 10% of cases [71]. This further supports the confounding nature of diagnosing DILI in patients with underlying autoimmune liver diseases.

2.5. DILI In Individuals With Cholestatic Liver Disease

More recently, the differential risk of DILI based on the underlying aetiology of the liver disease, independent of hepatic impairment, became evident when concurrent trials in MASH and PBC were conducted using the same daily dosage [72, 73, 74, 75, 76, 77, 78, 79, 80, 81]. These trials showed unexpected cases of elevated liver enzymes in the PBC but not in the MASH clinical trials [77, 81, 82]. This raised concern for an increased risk of DILI, leading to lower daily dosage in subsequent PBC clinical trials.

A few drugs initially developed for MASH are also under evaluation for treating PBC due to shared mechanisms related to liver inflammation, fibrosis and metabolic pathways [73, 76, 77, 78, 80]. These first‐in‐class, novel therapeutic agents were predominantly agonists of nuclear receptors targeting peroxisome proliferator‐activated receptor (PPAR), farnesoid X receptor, or fibroblast growth factor 19, and apical bile acid transport inhibitors [74, 77, 78, 80, 81, 83, 84]. Early dose‐finding clinical trials revealed an unexpectedly higher incidence of liver safety events in PBC patients at higher daily dosages [77, 82]. First was the OCA trial that reported one subject in the 10 mg daily arm vs. three subjects in the 50 mg daily group with elevated bilirubin or aminotransferase levels [85]. Next was the seladelpar PBC trial that reported three subjects treated with seladelpar (one on 50 mg, two on 200 mg) who developed fully reversible, asymptomatic liver enzyme elevations (5–20 × ULN), leading to early termination of the study [82]. These elevations were attributed to a dose‐dependent risk of hepatotoxicity in PBC as concurrently conducted trials in non‐cirrhotic MASH with obeticholic acid (10 and 25 mg) and seladelpar (10, 20 or 50 mg once daily) did not raise any concern for risk of DILI.

In the seladelpar PBC program, a phase 2 trial revealed one patient with increases in ALT and AST levels (Grades 2 and 3, respectively, concomitant with rifampicin use and adjudicated as possibly related to either seladelpar or rifampicin) [86]. Two cases with transient increases in bilirubin were deemed unrelated to seladelpar. Of these, one had increased ALT concomitant with worsening of rheumatoid arthritis and use of ibuprofen. The two patients who experienced ALT and AST elevations were on concomitant rifampicin administration for pruritus. Interestingly, both these patients with ALT/AST elevations were cirrhotic at baseline; one discontinued the study due to mild AST elevations after 23 weeks of treatment, and the other completed the study after aminotransferase elevations (> 3 × ULN) resolved [86]. However, no cases of DILI due to seladelpar were reported in the phase 3 trials [80, 87]. Finally, in the seladelpar open‐label extension study, there was one case of elevated bilirubin levels that met the study liver safety monitoring criteria (> 1.5 × baseline value), which was attributed to the progression of PBC (severe ductopenia noted on a post‐treatment biopsy) [88]. In another patient, periodic increases in liver tests (Grade 2 elevations in bilirubin and AST) with a temporal relationship with rheumatoid arthritis flares and increased use of non‐steroidal anti‐inflammatory drugs, which resolved upon discontinuation of seladelpar, was considered possibly related to seladelpar [88].

In the elafibranor program, similar doses (80 and 120 mg daily) were explored for non‐cirrhotic MASH and PBC. More than 2000 patients with MASH were exposed to elafibranor for nearly 72 weeks in the phase 3 trial, and no cases of hepatotoxicity were reported [89]. However, in the phase 3 PBC trial, one possible case of DILI was reported out of 108 subjects in the elafibranor group [90]. Specifically, one patient in the elafibranor group (0.9%) and two patients (causality deemed probable) in the placebo group (3.8%) experienced elevated aminotransferase levels (> 3 × ULN or baseline) or elevated bilirubin levels (> 2 × ULN), or both, meeting the protocol‐defined thresholds [90].

Lastly, in the saroglitazar PBC phase 2 trial, two patients in the 4 mg daily dosage group experienced significant increases in aminotransferase (> 5 × baseline), adjudicated as ‘highly likely’ and ‘probable’ for causality [77]. Another subject in this group had aminotransferases > 3 × baseline, adjudicated as ‘probable.’ One subject in the saroglitazar 2 mg group had an ALT increase of > 2× baseline value, also deemed ‘probable.’ There were no cases that met Hy's law criteria, and there were no increases in alkaline phosphatase or bilirubin, nor were any new symptoms reported [77]. Liver enzyme levels returned to normal within 3 months in all subjects [77]. No increases in the liver enzymes were reported in the non‐cirrhotic MASH trials despite the fact that the number of subjects exposed was much higher [81].

One nuance about DILI with PPARs in patients with PBC is the possibility of drug‐induced autoimmune‐like hepatitis. A new diagnosis of overlap syndrome occurred in PBC patients while participating in clinical trials with bezafibrate and, more recently, with saroglitazar [77]. In the bezafibrate phase 3 trial, three patients developed aminotransferase elevations that were 5 × ULN, leading to permanent discontinuation of the drug in two of them [91]. Interestingly, two of the three patients required glucocorticoid administration for suspected overlap syndrome with autoimmune hepatitis [91]. In the saroglitazar phase 2 trial, one subject in the 4 mg daily dosage arm developed overlap syndrome and needed treatment with immunosuppressant treatment [77].

2.6. DILI In Individuals With Cholestatic Liver Disease and Hepatic Impairment

Currently, OCA, the first 2nd line agent approved in 2016, carries a boxed warning and is contraindicated in patients with decompensated cirrhosis (e.g. Child‐Pugh Class B or C), a prior decompensation event, compensated cirrhosis with evidence of portal hypertension (e.g. ascites, gastroesophageal varices, persistent thrombocytopenia), or complete biliary obstruction https://www.fda.gov/media/149516/download?attachment. Prior to these boxed warnings, OCA was used in patients with advanced liver disease and hepatic impairment, and soon after, there were reports of worsening liver disease (marked increase in bilirubin occurring several months after start of OCA), suggesting that it may have contributed to liver injury that could not be explained solely by the natural progression of disease [92]. Studies on hepatic impairment in subjects with mild, moderate, and severe conditions (Child‐Pugh Class A, B, and C) showed altered pharmacokinetics, suggesting that the risk of DILI may be related to these changes. The mean AUC of total OCA increased by 1.1‐fold in mild, 4‐fold in moderate, and 17‐fold in severe impairment compared to subjects with normal liver function after a single 10 mg dose of OCA https://www.accessdata.fda.gov/drugsatfda_docs/label/2022/207999s008lbl.pdf. The major active conjugates with substantially higher exposure than OCA due to the long half‐life, substantial exposure accumulation following multiple doses, and systemic exposure presented as total OCA concentration (= OCA + Tauro‐OCA + Glyco‐OCA). This accumulation may be further exacerbated in those with cholestatic cirrhosis with evidence of portal hypertension suggesting that alterations in PK and risk of DILI may be higher in this patient population for drugs that are hepatically metabolised. Therefore, we expect the AUCs of total OCA to be much higher than those reported in the approved package information since the hepatic impairment study did not include those with cholestatic cirrhosis.

Further insights into altered PK in cholestatic liver disease, both with and without cirrhosis and portal hypertension, were gained from the saroglitazar hepatic impairment PK studies [93, 94]. These studies showed no significant differences in drug exposure between patients with non‐cirrhotic PBC and matched controls with normal hepatic function [93, 94]. Similarly, minimal differences in exposure were observed in patients with mild hepatic impairment on days 1 and 28 when taking 1 or 2 mg daily [94]. In contrast, patients with moderate hepatic impairment experienced a significant increase in saroglitazar exposure [94]. Patients with moderate hepatic impairment had exposure levels almost 3‐fold higher in the 2 mg daily dosage arm compared to 1 mg daily dosage arm, suggesting that those with moderate hepatic impairment were at most risk of higher exposures at higher daily dosages [94]. In contrast, in the pharmacokinetic studies in MASH cirrhosis, saroglitazar exposure was minimally increased (< 20%) in those with mild and moderate hepatic impairment, both with and without portal hypertension [93]. The exposure increased only in subjects with severe hepatic impairment, nearly three times higher compared to individuals with normal hepatic function [93]. These findings suggest that drug exposure is significantly increased at higher dosages in PBC, even in patients with moderate hepatic impairment, unlike in MASH, where increased exposure is only observed in those with severe hepatic impairment.

2.7. Systemic Diseases With Liver Involvement That Mimic DILI

Some diseases have associated liver involvement and have elevated liver tests as part of their clinical presentation and may not meet the criteria for a diagnosis of CLD. Notable examples include tuberculosis, syphilis, sarcoidosis, lymphoma, and systemic mastocytosis. Although there is no conclusive evidence to suggest an increased risk of DILI from systemic disease involving the liver, there are some instances that indicate such a possibility. Early evidence for it was observed in human immunodeficiency virus (HIV) patients, who are noted to be at a higher risk for DILI from trimethoprim‐sulfamethoxazole (TMP‐SMX) compared to HIV‐negative individuals [95]. Furthermore, there is some suggestion that the DILI episode occurs at the time of immune reconstitution after initiation of antiretroviral therapy in AIDS patients.

3. Outcomes of DILI in Patients With Chronic Liver Disease

While the majority of patients experiencing DILI recover from their liver injury, this is sadly not universal. In the 899 cases of DILI studied in the DILIN 10% of patients sadly required liver transplantation (LT) due to liver failure or died [2], Additionally, 17% of patients experienced chronic DILI, defined in the US DILIN as persistent liver test abnormalities more than 6 months after DILI onset, which was more common in patients older than 65 and those with a cholestatic pattern of liver injury [2]. Approximately 13% of patients with DILI had persistent liver test elevations (defined as ALT or AST > 1.5 × ULN or alkaline phosphatase > 2 × ULN), although this was not associated with underlying CLD [96]. In the Spanish DILI registry, 8% of patients developed chronic DILI (defined as AST, ALT, bilirubin or alkaline phosphatase > 1 × ULN) but patients with underlying CLD were excluded in that analysis [97].

Beyond the risk of DILI associated with underlying CLD is the risk for worse outcomes following DILI. Approximately 10% of patients with DILI in this DILIN cohort had underlying CLD, and those patients experienced a 3‐fold higher rate of mortality or LT [2]. These findings are mirrored by the Spanish Registry for DILI, where the rate of underlying CLD among patients with DILI was 6.3%, with an associated 4‐fold increase in risk of liver‐related death [3]. Together, these data suggest that underlying CLD predisposes to increased frequency and severity of DILI.

An examination of 107 deaths occurring within 2 years of DILI among 1089 patients in the DILIN cohort indicated that DILI had a primary role in 68 (64%), a contributory role in 15 (14%), and no role in 22 (21%), while two cases could not be categorised [98]. When DILI had a primary role in death, it was associated with acute, acute cholestatic, chronic, and acute‐on‐chronic liver failure (ACLF) in 74%, 6%, 13%, and 7%, respectively [98]. A new R ratio for Hy's law (nR criteria for Hy's Law), incorporating the degree of AST and bilirubin elevation, and AST to ALT ratio at DILI onset, better predicted 6‐month mortality or LT than the original Hy's law [98, 99].

Patients with CLD and a superimposed DILI resulting in a new hepatic decompensation and possibly other organ failure would meet the definition of ACLF. Some definitions of ACLF are dependent on the presence of underlying cirrhosis for some definitions, for example, the seminal European and North American studies describing ACLF [100, 101]. Notably, cirrhosis is not a prerequisite in the Asia Pacific Association for the Study of the Liver [102], the World Gastroenterology Organisation [103], and the American College of Gastroenterology definitions of ACLF [104]. DILI accounts for a small proportion of ACLF globally (1%–10%) with geographic differences in underlying liver disease (a more common viral disease in the east) and a class of implicated agents (antituberculosis, HDS, or complementary and alternative medicine in the east), while mortality rates are uniformly high ranging from 35% to 50% [105, 106, 107, 108, 109, 110, 111]. Studies describing DILI related ACLF, and their associated outcomes are summarised in Table 1, which has been adapted from a recent review by Ma et al. [106].

TABLE 1.

Selected characteristics and features of studying evaluating drug induced acute on chronic liver failure.

Study	Location	Number of patients studied	Definition of ACLF Used	Proportion DI‐ACLF	Implicated Agents	Underlying Liver Disease	Mortality rates	Comment
Shalimar et al. [107]	India, single center	213	APASL	11/132, 5.2%	Antituberculosis drugs	40% alcohol, 24% HBV, 20% cryptogenic	In hospital: Overall: 43% AT drugs: 55%	DI‐ACLF due to antituberculosis drugs did not independently predict mortality
Shalimar et al. [108]	India, multicenter	1049 consecutive patients, 381 with complete data	APASL	17/381, 4.5%	Antituberculosis drugs, antiepileptics	52% alcohol, 21% cryptogenic, 16% viral	In hospital: Overall: 148/381, 39% Drug related: 6/17, 34%	DI‐ACLF did not independently predict mortality compared to other precipitants
Devarbhavi et al. [105]	Asia, multinational	3132	APASL	329/3132, 11%	Antituberculosis drugs, CAMs, methotrexate (n = 2), antiepileptic drug (n = 1)	29% alcohol, 26% cryptogenic, 17% HBV, 17% NASH	90‐day: Non‐DI‐ACLF: 39% DI‐ACLF: 47%	DI‐ACLF frequently presented with jaundice (100%), ascites (88%), and encephalopathy (47%) Arterial lactate and total bilirubin were independent predictors of mortality in DI‐ACLF
Shi et al. [112]	China, single center	322	EASL‐CLIF	10/322 (2.5%)	Not specified	64% HBV, 11% alcohol, 14% both	90‐day: Hepatic ACLF (including DI): 59% a Extrahepatic ACLF: 68% a	Extrahepatic ACLF had significantly higher 90 day and 1 year mortality versus hepatic ACLF
Li et al. [113]	China, single center	300	EASL‐CLIF	7/300 (2.3%)	Not specified	100% HBV	90‐day: ACLF: 50% a No ACLF: 4.6% a	The use of hepatotoxic drugs and herbs was not found to be an independent predictor of developing ACLF during hospitalisation. The overall rates of hepatotoxic drug or herb intake prior to admission did not differ significantly between patients presenting with ACLF versus no ACLF
Sun et al. [111]	China, multicenter	335	EASL‐CLIF	72/335 with chronic HBV had suspected DILI (21.4%) 260 developed ACLF, 45 with suspected DILI (17%)	Herbals (44%) statins (17%) Hypoglycemic agents (14%) Antituberculosis drugs (10%) Immunosuppressive agents (6%) antiepileptic drug (4%) NSAIDs (3%)	100% HBV, 76% had cirrhosis	90‐day: DILI: 54% Non‐DILI: 38%	There were higher grades of ACLF in patients with chronic HBV and suspected DILI compared with other etiologies. ACLF was not associated with HBV viral load or e antigen status
Maipang et al. [114]	Thailand, single center	343	EASL‐CLIF	3/343 (0.9%)	Not specified	38% HBV, 21% HCV, 20% cryptogenic	90‐day: Hepatic ACLF (including DI‐ACLF): 73% Extrahepatic ACLF: 59%	Hepatic ACLF had significantly higher 28‐day mortality versus extrahepatic ACLF. Mortality at 90‐day, 6 months, and 1‐year were similar between the two groups

Abbreviations: APASL, Asia‐Pacific Association for Study of the Liver; CAM, complementary and alternative medicine, DI‐ACLF, drug induced‐acute on chronic liver failure, EASL‐CLIF, European Association for Study of the Liver‐Chronic Liver Failure, HBV, hepatitis B virus, HCV, hepatitis c virus, NASH, nonalcoholic steatohepatitis.

^a Statistically significant.

Author Contributions

Study design: all authors. Data analysis: all authors. Manuscript drafting: all authors. Data interpretation and review and revision of the manuscript: all authors.

Conflicts of Interest

None of the authors have relevant disclosures with the following exceptions. Dr. Chalasani declares no COIs for this paper. For full disclosure, he has had paid consulting agreements with Madrigal, GSK, Zydus, Altimune, BioMea Fusion, Ipsen, Akero, Merck, and Pfizer. He has research grants from Boehringer‐Ingelheim and Exact Sciences. He has equity ownership in Avant Sante, a contract research organisation, and Heligenics, a drug discovery start‐up company. Dr. Vuppalanchi declares no COIs for this paper. For full disclosure, he receives institutional research grant support from Galectin Therapeutics, Gilead Sciences, Eli Lilly, Astra Zeneca, Takeda, GSK, Zydus Therapeutics Inc. Also, he discloses consulting for Fortrea, Medpace, WCG Clinical, Icon, Jazz Pharmaceuticals, Madrigal, Hightide, Intercept, Akero, Cymabay/Gilead, Hanmi, Merck, and GSK. Dr. Ghabril declares no COIs for this paper. For full disclosure, he receives institutional research grant support from Salix/ Bausch. He discloses consulting for CymaBay/Gilead, Zydus Therapeutics Inc., and BioCryst.

Data Availability Statement

Data sharing is not applicable to this article as no new data were created or analysed in this study.